swactor/crates/iroh-driver/src/telemetry_transport.rs

541 lines
18 KiB
Rust
Raw Normal View History

//! Iroh/QUIC transport adapter for telemetry subscriptions.
use std::error::Error;
use std::sync::Arc;
use std::time::Duration;
use crossbeam_channel::TryRecvError;
use iroh::endpoint::{Connection, RecvStream, SendStream};
use iroh::{Endpoint, EndpointAddr};
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
use swactor_engine::EngineHandle;
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
use telemetry::frame::{
ChannelDescriptor, ChannelId, ChannelRef, FrameDelivery, Position, StreamDescriptor,
TelemetryEvent,
};
use telemetry::{TelemetrySnapshot, TelemetrySubscription};
pub const TELEMETRY_ALPN: &[u8] = b"swactor/telemetry/0";
const MAGIC: &[u8; 4] = b"DSQ1";
const TAG_CHANNEL_DECLARED: u8 = 0x01;
const TAG_FRAME: u8 = 0x02;
const TAG_STREAM_ENDED: u8 = 0x03;
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
// ─── Pull model: collector-initiated subscriptions ───────────────────────────
//
// A supervising node dials a freshly-bootstrapped node on `TELEMETRY_ALPN`,
// sends one subscription request on the first uni stream, and the node answers
// by writing the existing header+events stream shape on a uni stream of the
// same connection. Subscription lifetime = connection lifetime.
/// Write a pull request (magic + flow id + token + `SubscriptionRequest`) and
/// finish the stream so the serving side's read completes.
const REQUEST_MAGIC: &[u8; 4] = b"DSQR";
pub async fn write_pull_request(
send: &mut SendStream,
flow_id: [u8; 16],
token: &[u8],
request: &telemetry::SubscriptionRequest,
) -> Result<(), BoxError> {
if token.len() > u16::MAX as usize {
return Err("telemetry pull token exceeds u16 length prefix".into());
}
send.write_all(REQUEST_MAGIC).await?;
send.write_all(&flow_id).await?;
send.write_all(&(token.len() as u16).to_le_bytes()).await?;
send.write_all(token).await?;
write_json(send, request).await?;
send.finish()?;
Ok(())
}
/// Read a pull request written by [`write_pull_request`].
pub async fn read_pull_request(
recv: &mut RecvStream,
) -> Result<([u8; 16], Vec<u8>, telemetry::SubscriptionRequest), BoxError> {
let mut magic = [0u8; 4];
recv.read_exact(&mut magic).await?;
if &magic != REQUEST_MAGIC {
return Err("invalid telemetry pull request magic".into());
}
let mut flow_id = [0u8; 16];
recv.read_exact(&mut flow_id).await?;
let mut token_len = [0u8; 2];
recv.read_exact(&mut token_len).await?;
let token_len = u16::from_le_bytes(token_len) as usize;
let mut token = vec![0u8; token_len];
recv.read_exact(&mut token).await?;
let request = read_json(recv).await?;
Ok((flow_id, token, request))
}
/// Node side: serve one accepted `TELEMETRY_ALPN` connection. Reads the pull
/// request from the first uni stream, subscribes the local endpoint, and
/// writes the answering subscription stream (header + events, until the
/// subscription ends or the connection drops) on a uni stream of the same
/// connection. One request per connection; the task exits when the writer
/// ends or the connection fails.
pub fn spawn_pull_server(
engine: &EngineHandle,
conn: Connection,
endpoint: std::sync::Arc<telemetry::TelemetryEndpoint>,
idle_sleep: Duration,
) {
let engine_handle = engine.clone();
engine.spawn(async move {
let Ok(mut recv) = conn.accept_uni().await else {
return;
};
let Ok((_flow_id, token, request)) = read_pull_request(&mut recv).await else {
return;
};
let subscription = endpoint.subscribe("supervisor-pull", request);
let header =
match TelemetryQuicHeader::from_snapshot(_flow_id, token, subscription.snapshot()) {
Ok(header) => header,
Err(_) => return,
};
let Ok(send) = conn.open_uni().await else {
return;
};
let _ =
write_subscription_until_closed(&engine_handle, send, header, subscription, idle_sleep)
.await;
});
}
/// Supervisor side: retain a pull subscription to a node on `TELEMETRY_ALPN`.
///
/// A transport interruption reconnects with bounded backoff. Returning after
/// the first EOF leaves a healthy node permanently stale, which is especially
/// easy to trigger while several freshly-bootstrapped nodes answer at once.
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
pub fn spawn_pull_collector(
engine: &EngineHandle,
endpoint: Endpoint,
peer: EndpointAddr,
flow_id: [u8; 16],
token: Vec<u8>,
request: telemetry::SubscriptionRequest,
fanout: std::sync::Arc<telemetry::DeliveryFanout>,
on_header: std::sync::mpsc::Sender<TelemetryQuicHeader>,
) {
let engine_handle = engine.clone();
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
engine.spawn(async move {
let peer_id = peer.id.to_string();
let mut retry_delay = Duration::from_millis(250);
loop {
match collect_pull_once(
&endpoint, &peer, flow_id, &token, &request, &fanout, &on_header,
)
.await
{
Ok(()) => return,
Err(error) => {
eprintln!(
"telemetry-pull: {peer_id}: {error}; retrying in {} ms",
retry_delay.as_millis()
);
}
}
engine_handle.timer(retry_delay).await;
retry_delay = retry_delay
.checked_mul(2)
.unwrap_or(Duration::from_secs(5))
.min(Duration::from_secs(5));
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
}
});
}
async fn collect_pull_once(
endpoint: &Endpoint,
peer: &EndpointAddr,
flow_id: [u8; 16],
token: &[u8],
request: &telemetry::SubscriptionRequest,
fanout: &telemetry::DeliveryFanout,
on_header: &std::sync::mpsc::Sender<TelemetryQuicHeader>,
) -> Result<(), String> {
let conn = endpoint
.connect(peer.clone(), TELEMETRY_ALPN)
.await
.map_err(|error| format!("connect failed: {error}"))?;
let mut req = conn
.open_uni()
.await
.map_err(|error| format!("open request stream failed: {error}"))?;
write_pull_request(&mut req, flow_id, token, request)
.await
.map_err(|error| format!("write request failed: {error}"))?;
let mut recv = conn
.accept_uni()
.await
.map_err(|error| format!("no answer stream: {error}"))?;
let header = read_header(&mut recv)
.await
.map_err(|error| format!("answer header unreadable: {error}"))?;
if on_header.send(header.clone()).is_err() {
return Ok(());
}
let stream = header.stream;
loop {
match read_next_event(&mut recv, &stream).await {
Ok(Some(event)) => {
fanout.publish(event);
}
Ok(None) => return Err("answer stream closed".to_owned()),
Err(error) => return Err(format!("read answer stream failed: {error}")),
}
}
}
const MAX_RECORD_BYTES: usize = 16 * 1024 * 1024;
type BoxError = Box<dyn Error + Send + Sync + 'static>;
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct TelemetryQuicHeader {
pub flow_id: [u8; 16],
pub token: Vec<u8>,
pub stream: StreamDescriptor,
pub channels: Vec<ChannelDescriptor>,
}
impl TelemetryQuicHeader {
pub fn new(
flow_id: [u8; 16],
token: impl Into<Vec<u8>>,
stream: StreamDescriptor,
channels: Vec<ChannelDescriptor>,
) -> Self {
Self {
flow_id,
token: token.into(),
stream,
channels,
}
}
pub fn from_snapshot(
flow_id: [u8; 16],
token: impl Into<Vec<u8>>,
snapshot: &TelemetrySnapshot,
) -> Result<Self, BoxError> {
let stream = snapshot
.streams
.first()
.cloned()
.ok_or("telemetry subscription snapshot has no stream")?;
Ok(Self::new(flow_id, token, stream, snapshot.channels.clone()))
}
}
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub struct TelemetryQuicWriteStats {
pub events: usize,
pub bytes: usize,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct TelemetryQuicRead {
pub header: TelemetryQuicHeader,
pub events: Vec<TelemetryEvent>,
}
pub fn spawn_subscription_writer(
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine: &EngineHandle,
endpoint: Endpoint,
peer: EndpointAddr,
header: TelemetryQuicHeader,
subscription: TelemetrySubscription,
idle_sleep: Duration,
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
) {
let engine_handle = engine.clone();
engine.spawn(async move {
let Ok(conn) = endpoint.connect(peer, TELEMETRY_ALPN).await else {
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
return;
};
let Ok(send) = conn.open_uni().await else {
return;
};
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
let _ =
write_subscription_until_closed(&engine_handle, send, header, subscription, idle_sleep)
.await;
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
});
}
pub async fn write_available_subscription(
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine: &EngineHandle,
send: SendStream,
header: &TelemetryQuicHeader,
subscription: &TelemetrySubscription,
) -> Result<TelemetryQuicWriteStats, BoxError> {
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
write_subscription_inner(engine, send, header, subscription, None).await
}
pub async fn write_subscription_until_closed(
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine: &EngineHandle,
mut send: SendStream,
header: TelemetryQuicHeader,
subscription: TelemetrySubscription,
idle_sleep: Duration,
) -> Result<TelemetryQuicWriteStats, BoxError> {
write_header(&mut send, &header).await?;
let mut stats = TelemetryQuicWriteStats::default();
loop {
match subscription.try_recv() {
Ok(event) => {
let bytes = write_event(&mut send, &event).await?;
if bytes > 0 {
stats.bytes += bytes;
stats.events += 1;
}
}
Err(TryRecvError::Empty) => {
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine.timer(idle_sleep).await;
}
Err(TryRecvError::Disconnected) => break,
}
}
send.finish()?;
Ok(stats)
}
async fn write_subscription_inner(
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine: &EngineHandle,
mut send: SendStream,
header: &TelemetryQuicHeader,
subscription: &TelemetrySubscription,
idle_sleep: Option<Duration>,
) -> Result<TelemetryQuicWriteStats, BoxError> {
write_header(&mut send, header).await?;
let mut stats = TelemetryQuicWriteStats::default();
loop {
match subscription.try_recv() {
Ok(event) => {
let bytes = write_event(&mut send, &event).await?;
if bytes > 0 {
stats.bytes += bytes;
stats.events += 1;
}
}
Err(TryRecvError::Empty) => match idle_sleep {
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
Some(delay) => engine.timer(delay).await,
None => break,
},
Err(TryRecvError::Disconnected) => break,
}
}
send.finish()?;
Ok(stats)
}
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
pub async fn write_event(send: &mut SendStream, event: &TelemetryEvent) -> Result<usize, BoxError> {
let mut bytes = Vec::new();
match event {
TelemetryEvent::StreamDeclared(_) => return Ok(0),
TelemetryEvent::ChannelDeclared(descriptor) => {
bytes.push(TAG_CHANNEL_DECLARED);
put_json(&mut bytes, descriptor)?;
}
TelemetryEvent::Frame(delivery) => {
bytes.push(TAG_FRAME);
bytes.extend_from_slice(&delivery.channel.channel.0.to_le_bytes());
bytes.extend_from_slice(&delivery.position.0.to_le_bytes());
put_bytes(&mut bytes, &delivery.payload)?;
}
TelemetryEvent::StreamEnded(_) => {
bytes.push(TAG_STREAM_ENDED);
}
}
if bytes.len() > MAX_RECORD_BYTES {
return Err("telemetry QUIC record exceeds max size".into());
}
send.write_all(&(bytes.len() as u32).to_le_bytes()).await?;
send.write_all(&bytes).await?;
Ok(4 + bytes.len())
}
pub async fn read_stream_header(recv: &mut RecvStream) -> Result<TelemetryQuicHeader, BoxError> {
read_header(recv).await
}
pub async fn read_events_from_stream(mut recv: RecvStream) -> Result<TelemetryQuicRead, BoxError> {
let header = read_header(&mut recv).await?;
let mut events = Vec::new();
loop {
match read_next_event(&mut recv, &header.stream).await? {
Some(event) => events.push(event),
None => break,
}
}
Ok(TelemetryQuicRead { header, events })
}
pub async fn read_next_uni_from_connection(
conn: &Connection,
) -> Result<TelemetryQuicRead, BoxError> {
let recv = conn.accept_uni().await?;
read_events_from_stream(recv).await
}
pub fn spawn_connection_reader(
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
engine: &EngineHandle,
conn: Connection,
sink: std::sync::mpsc::Sender<TelemetryEvent>,
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
) {
engine.spawn(async move {
loop {
let recv = match conn.accept_uni().await {
Ok(recv) => recv,
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
Err(_) => return,
};
let Ok(read) = read_events_from_stream(recv).await else {
continue;
};
for event in read.events {
if sink.send(event).is_err() {
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
return;
}
}
}
feat(engine): substrate-neutral execution engine abstraction Introduce the swactor engine: a swactor-owned composite that retains a selected execution substrate, drives the core runtime, and hosts the async/blocking/timer work that backs actors. Integrations receive one cloneable EngineHandle and never construct or borrow a raw Tokio runtime/handle. Engine crate (crates/engine): - The contract: spawn / spawn_blocking / timer / interval / now, a per-implementation capability model with construction-time binding (require()), and engine-owned time. The engine owns all progression; actor handlers stay synchronous and never .await. - TokioBackend owns the Tokio runtime and schedules core ticks and supporting futures on it; SteppingBackend is a single-threaded deterministic scheduler with virtual time (the non-Tokio portability proof). Core is driven through its existing tick() surface; a self-rescheduling CoreDriver is installed at construction and is the sole place permitted to call try_tick. iroh-driver: - Receives an EngineHandle instead of a raw Tokio Handle. Accepts, reads, dials, writes, endpoint construction, and teardown schedule through it; required capabilities (tasks/timers/io) are validated before the endpoint binds. Engine-hosted interval pumps drive actor-bridge, datastream, and edge ingress. myelin: - One node/orchestrator engine owns core, protocol tick injection, and transport progression; the application loop only drains integration-owned queues. Stage-shard process readers, delayed actor messages, helper stdout/stderr, prompt RPC, and CPU sampling all schedule through the engine (spawn_blocking / engine tasks / timers). - Removed the split-engine APIs: install_actor_bridge_pump(period) and spawn_protocol_ticker(period) use each component's stored engine; deleted the no-op pump_network callback and its plumbing; deleted the dashboard raw-Tokio/standalone-runtime conveniences. Enforcement: - A clippy disallowed-methods boundary forbids direct runtime/scheduling/ time/core-driving bypasses, denied in swactor-engine, iroh-driver, and myelin. Retained excluded uses (VastAI provider, provider process supervision/log capture, OS-signal/stdin/process-control sequencing) carry narrow allowances with reasons. Verification: - Engine contract + unit tests (incl. the SteppingBackend portability proof), iroh integration tests (capability rejection before binding, multi-node actor behavior), and a production execution-composition smoke test that observes engine-driven actor progress with no ambient Tokio runtime and no manual tick/pump. Workspace all-target/all-feature clippy and tests are green. Specs co-located with their crates: ENGINE_SPEC.md in crates/engine, IROH_DRIVER_SPEC.md in crates/iroh-driver. VastAI remains explicitly out of scope pending its separate redesign.
2026-08-10 20:23:03 +00:00
});
}
provisioning-reconciler-demo: wire-announce readiness + --docker node kind Replace the per-attempt key-file side channel (DEMO_NODE_KEY_FILE, read_key_report, JoinCheck) with a control-plane announce: node roles send a tagged gossip frame {attempt, logical_node, key_hex, endpoint_addr_json} after joining and every heartbeat thereafter. - iroh-driver: IrohDriver::send_tagged_gossip — fire-and-forget tag-routed gossip egress for bridge-less clients (reuses cached/join connections, dials with backoff). - provisioning: BootstrapMsg::Announce — first delivery while bootstrapping completes the attempt (collector + exactly-once Bootstrapped report); duplicates, misrouted attempts, and terminal-phase announces drop. Unit-tested. - xtask demo: AnnounceActor decodes the tag-routed frame and forwards by attempt to the owning bootstrap actor; last_announce_ms is the wire heartbeat. LocalProcessLogic keeps only process lifecycle. - --docker: DockerProcessLogic (kind "docker") — attached "docker run --rm" child on a per-run labeled bridge network (foreign-node masking: per-container IPs, gateway-dialed supervisor). Standalone scratch image from the static-musl xtask binary (37MB), staged one-file build context. The container is force-removed on every terminal path so a SIGKILLed docker CLI cannot orphan a running container. - Cleanup: no volumes/mounts; label-filtered exit sweep (SIGINT and SIGTERM both drain first) + startup sweep of stale demo resources; images persist per run token. Verified live: process kind (kill -> replacement in 3.4s, provision/ remove/kill waves, zero orphans) and docker kind (8-node abuse waves across docker kill, mid-provision control kills, CLI SIGKILL orphans force-removed, SIGKILL-crash leftovers swept on restart, clean exits leave zero containers/networks/CLIs). provisioning 22 + iroh-driver 13 tests pass.
2026-08-16 16:30:52 +00:00
async fn write_header(send: &mut SendStream, header: &TelemetryQuicHeader) -> Result<(), BoxError> {
if header.token.len() > u16::MAX as usize {
return Err("telemetry token exceeds u16 length prefix".into());
}
send.write_all(MAGIC).await?;
send.write_all(&header.flow_id).await?;
send.write_all(&(header.token.len() as u16).to_le_bytes())
.await?;
send.write_all(&header.token).await?;
write_json(send, &header.stream).await?;
write_json(send, &header.channels).await?;
Ok(())
}
async fn read_header(recv: &mut RecvStream) -> Result<TelemetryQuicHeader, BoxError> {
let mut magic = [0u8; 4];
recv.read_exact(&mut magic).await?;
if &magic != MAGIC {
return Err("invalid telemetry QUIC magic".into());
}
let mut flow_id = [0u8; 16];
recv.read_exact(&mut flow_id).await?;
let mut token_len = [0u8; 2];
recv.read_exact(&mut token_len).await?;
let token_len = u16::from_le_bytes(token_len) as usize;
let mut token = vec![0u8; token_len];
recv.read_exact(&mut token).await?;
let stream = read_json(recv).await?;
let channels = read_json(recv).await?;
Ok(TelemetryQuicHeader {
flow_id,
token,
stream,
channels,
})
}
pub async fn read_next_event(
recv: &mut RecvStream,
stream: &StreamDescriptor,
) -> Result<Option<TelemetryEvent>, BoxError> {
let mut len = [0u8; 4];
if recv.read_exact(&mut len).await.is_err() {
return Ok(None);
}
let len = u32::from_le_bytes(len) as usize;
if len > MAX_RECORD_BYTES {
return Err("telemetry QUIC record exceeds max size".into());
}
let mut buf = vec![0u8; len];
recv.read_exact(&mut buf).await?;
decode_record(&buf, stream).map(Some)
}
fn decode_record(buf: &[u8], stream: &StreamDescriptor) -> Result<TelemetryEvent, BoxError> {
if buf.is_empty() {
return Err("empty telemetry QUIC record".into());
}
match buf[0] {
TAG_CHANNEL_DECLARED => {
let descriptor: ChannelDescriptor = serde_json::from_slice(&buf[1..])?;
Ok(TelemetryEvent::ChannelDeclared(descriptor))
}
TAG_FRAME => {
if buf.len() < 1 + 4 + 8 + 4 {
return Err("telemetry QUIC frame record truncated".into());
}
let channel = ChannelId(u32::from_le_bytes([buf[1], buf[2], buf[3], buf[4]]));
let position = Position(u64::from_le_bytes([
buf[5], buf[6], buf[7], buf[8], buf[9], buf[10], buf[11], buf[12],
]));
let mut len = [0u8; 4];
len.copy_from_slice(&buf[13..17]);
let payload_len = u32::from_le_bytes(len) as usize;
let payload = buf
.get(17..17 + payload_len)
.ok_or("telemetry QUIC frame payload truncated")?;
if 17 + payload_len != buf.len() {
return Err("bytes remain after telemetry QUIC frame record".into());
}
Ok(TelemetryEvent::Frame(FrameDelivery {
channel: ChannelRef {
stream: stream.stream.clone(),
channel,
},
position,
payload: payload.to_vec(),
}))
}
TAG_STREAM_ENDED => Ok(TelemetryEvent::StreamEnded(stream.stream.clone())),
_ => Err("unknown telemetry QUIC record tag".into()),
}
}
pub async fn read_stream_into_fanout(
recv: RecvStream,
fanout: Arc<telemetry::DeliveryFanout>,
) -> Result<TelemetryQuicHeader, BoxError> {
let read = read_events_from_stream(recv).await?;
fanout.publish_batch(read.events);
Ok(read.header)
}
fn put_json<T: serde::Serialize>(out: &mut Vec<u8>, value: &T) -> Result<(), BoxError> {
out.extend_from_slice(&serde_json::to_vec(value)?);
Ok(())
}
fn put_bytes(out: &mut Vec<u8>, bytes: &[u8]) -> Result<(), BoxError> {
if bytes.len() > u32::MAX as usize {
return Err("telemetry delivery exceeds u32 length prefix".into());
}
out.extend_from_slice(&(bytes.len() as u32).to_le_bytes());
out.extend_from_slice(bytes);
Ok(())
}
async fn write_json<T: serde::Serialize>(send: &mut SendStream, value: &T) -> Result<(), BoxError> {
let bytes = serde_json::to_vec(value)?;
if bytes.len() > u32::MAX as usize {
return Err("telemetry header JSON exceeds u32 length prefix".into());
}
send.write_all(&(bytes.len() as u32).to_le_bytes()).await?;
send.write_all(&bytes).await?;
Ok(())
}
async fn read_json<T: serde::de::DeserializeOwned>(recv: &mut RecvStream) -> Result<T, BoxError> {
let mut len = [0u8; 4];
recv.read_exact(&mut len).await?;
let len = u32::from_le_bytes(len) as usize;
if len > MAX_RECORD_BYTES {
return Err("telemetry QUIC record exceeds max size".into());
}
let mut bytes = vec![0u8; len];
recv.read_exact(&mut bytes).await?;
Ok(serde_json::from_slice(&bytes)?)
}